Meta AI restructuring: why the 60% team-cut plan fell apart
The Meta AI restructuring, code-named Project OT, considered cutting staffing on many teams by up to 60%, according to internal planning documents. The two-wave effort built around "AI-native" operations envisioned a smaller and more "talent-dense" workforce at the owner of Facebook and Instagram, with small human teams supervising fleets of AI agents. The second wave never ran. Details of the plan, its partial execution, and its collapse emerged this week.
Under the plan, employees would either transfer to newly formed units or lose their jobs. One internal projection put the scale of job losses at a level equal to or greater than the roughly 25% headcount reduction Meta made three years earlier. The intended structure was flatter: smaller human teams aided by AI "virtual workers" would absorb the day-to-day workload of larger staff groups while a small number of people handled oversight.
Inside the Meta AI restructuring
The first wave landed in May. Meta notified roughly 8,000 employees, about 10% of its global workforce, that their roles were gone, starting with 4 a.m. emails in Singapore and rolling westward through the UK and the US. Roughly 7,000 employees were reassigned into AI-focused roles at the same time, and about 6,000 open positions were scrapped before they could be filled. Zuckerberg announced the round in a May 20 memo to staff that cautioned against assuming success.
The May cut came on top of earlier reductions. Headcount stood at 76,834 at the end of 2025, then fell about 3% to roughly 74,700 by the end of March after a February round that redirected compensation spending toward Meta Superintelligence Labs and core AI infrastructure. Henry Chen, a tech lead who had spent 13 years at Meta, was among the May layoffs, one of many veterans caught in the pivot to AI-first operations. Meta offered affected staff a severance package it described as generous, including up to 16 weeks of pay.
The restructuring unfolded in three rounds, each with a different outcome:
| Round | Timing | Scope | Status |
|---|---|---|---|
| February reduction | Q1 2026 | Roughly 3% of headcount | Completed |
| First wave | May 20, 2026 | 8,000 laid off (~10%), 7,000 reassigned | Completed |
| Second wave | Planned for 2026 | Up to 60% of some teams | Canceled |
Why the second wave collapsed
The planned second round never happened. Zuckerberg called it off after staff pushed back on automation-driven job cuts and challenged the productivity gains used to justify them. The technology was already trailing the plan. At an internal all-hands on July 2, Zuckerberg told staff that the agent work underpinning the reorganization had slowed, and he said he expects no further companywide layoffs for the rest of 2026. He later addressed the early-morning notification process during the company's earnings call, detailing the financial impact of the round.
The financial footprint of the Meta AI restructuring is already visible in the accounts. Meta booked $1.18 billion in severance costs in the second quarter, compressing operating margins at a company funding one of the largest AI infrastructure programs in the industry. Meta now expects its 2026 AI infrastructure spending to land between $130 billion and $145 billion. That is nearly twice 2025 spending and above the $125 billion guidance the company issued three months earlier. Investors have shown frustration with the spending plans, and the stock has fallen since the higher targets were announced.
Meta framed the May cuts as necessary to fund its push into artificial intelligence, the same push that now carries the $130 billion to $145 billion infrastructure price tag for 2026. That framing made the reversal harder to absorb: the headcount plan and the spending plan were presented as a single program, and only the first part has been publicly walked back.
The reorganization reached beyond Meta's own payroll. IT services outsourced to Wipro were cut by at least 25% after Meta closed its digital marketing division, and Wipro now expects roughly $75 million in annual revenue from the relationship. Zuckerberg has framed AI as a net job creator, arguing that the infrastructure needed to power the technology generates employment; those jobs sit in data-center construction, a different labor market from the corporate functions the restructuring eliminated. Construction and specialized data-center labor are becoming AI's clearest employment effect even as thousands of corporate roles disappear.
The abandoned plan also invites comparison with Meta's last major restructuring. Three years ago the company cut roughly a quarter of its staff in a round that remains a benchmark for its willingness to reduce headcount. The projection that the 2026 plan could remove headcount at a scale comparable to or greater than that round shows how aggressive it was. The difference is that this time the cuts were tied to an automation bet that has since slowed. That earlier round predated the agent era; Project OT assumed the agents were ready, and the July admission indicated they were not.
What collapsed with the Meta AI restructuring was a headcount plan and the productivity assumptions behind it. The design assumed agents could absorb a majority of the work on many teams, and the July admission that agent work had slowed undercut that assumption inside the company. The sequence is worth pausing on: Meta executed the first wave before its agents had demonstrated they could carry the workload, then canceled the second when the gap between projection and execution became visible. The severance bill, the margin pressure, and the public reversal together price the cost of overestimating AI productivity. The canceled wave is also a rare public reversal for a CEO who has otherwise pressed the AI agenda at full speed.
Set against a $130 billion-plus infrastructure budget, the $1.18 billion severance charge is a small fraction of the total. The restructuring was about reshaping the workforce to match a vision of AI-native operations, and that vision did not hold up against employee resistance and unproven agents. The company's public commitment to no further companywide layoffs in 2026 now caps how far the model can be pushed in the near term. For the employees moved into AI roles the reassignment stands, and for the teams that kept their headcount the agent takeover is deferred for now: the direction of travel is unchanged, with capital still flowing into AI infrastructure while the headcount plan waits for the agents to catch up.
The spending numbers put the episode in context. The 2026 infrastructure plan was increased within three months of the earlier guidance, and investors remain split on whether bills of that size will translate into returns. The failed restructuring adds a labor-cost data point to that debate, alongside the severance charge and the canceled second wave.
Why this matters
The episode is the clearest evidence yet that AI-productivity projections at the largest AI spenders can outrun what execution, employees, and even the CEO will tolerate. A company on track to spend $130 billion to $145 billion on AI infrastructure this year has shown that its internal productivity math did not hold up against its own workforce, and that employee resistance can force a retreat even when the CEO backs the plan. For teams planning similar AI-native restructures, the sequence matters: agents must carry the work before headcount is removed, and the cost of getting that order wrong is measured in severance, margins, and abandoned waves.
Related Articles
- Meta AI agent stall: Zuckerberg admits four-month lag
- Meta AI Workforce Restructuring: 8,000 Layoffs as 7,000 Staff Pivot to Agentic Teams
- Meta Realigns Workforce with $135 Billion AI Infrastructure Push
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.